There is a lot going on with AI now. It would be pretty cool to get a vector database on Cloudflare. https://github.com/qdrant/qdrant is one written in Rust. I'm not sure if it could be made to work with WebAssembly and DO Storage. I know myself and one other developer are working on AI products that need a vector database and we're using Cloudflare Durable Objects in our solutions.

Being able to deploy to a specific region (=> could allow us to do progressive deployments, and test new release on a subset of users easily)

Like this? https://developers.cloudflare.com/workers/examples/ab-testing/

Multiple named alarms for DO would be very handy

the ability to write worker data to local disk (up to the POST data limit), and read again (data deleted automatically after request is finished). i'm sure this could be useful for a number of things, but this is what got me thinking about this: i need to transform post data on the server before sending it to R2, which I can do using TransformStream, but then I cannot know the checksum to send to R2 ahead of time. If I could write to disk, I could stream it to the disk, calculate the checksum along the way, and then stream it from disk to R2.

You could put to R2, calculate the hash, then move it with the new hash in place?

but r2 has a hash option to verify the data integrity. i would need to transfer to r2, calculate the hash along the way, then read from r2 to compare hashes. though, it appears r2 calculates an md5 hash if none is supplied. so, it appears i could transfer to r2, calculate an md5 along the way, read the r2 generated hash and compare

still, i think a local disk cache could be a useful tool to offset the memory limit

The issue with a local cache is that edge blades have a surprisingly small amount of storage that they can just use for random tasks. There just isn’t disk space for anyone to just store a 1 GB file as a buffer

If you do need a buffer, you can use R2, KV, Cache(within cache limits), or DO storage

i was thinking more along the lines of 100mb. i understand there isn't enough disk space at the moment, but it might be something to look into going forward...maybe. i don't absolutely need it, and maybe never will, but it might be something to keep in mind as a potential nifty solution to other/future edge cases

C o n s i s t e n t dev envs across miniflare, wrangler dev, worker-rs -- I find myself having to relearn development frameworks everytime I change my project

If they arent consistent its just bugs that can be fixed so create issues on the respective repos. I have personally never had any issues between miniflare local, wrangler dev remote, or published 🤷

tbh I'm just not sure what are bugs and what are features. some examples: - you can define

env_path

in

[miniflare]

-- you can't do that in

wrangler dev --env=dev

- If you create

dev.var

in a worker-rs project, it gets picked up and the vars accessible from

ctx.secrets

when you run

wrangler dev

. You can't do the same thing in a JS worker project (or at the very least, my project doesn't seem to pick it up

It should be

.dev.vars

for all. Miniflare used to be a non cloudflare project so it having an option that others dont isnt too surprising.

okie i was being silly here -- ty ty

that actually explains alot -- seems like wrangler and miniflare have over-lapping responsibilities

Afaik they are working on integrating miniflare into wrangler directly (which I think is the experimental local mode?) Not too sure though I dont follow those news too much 😅

wrangler dev

has been running on miniflare for a while now.

wrangler

just providers a more ergonomic way to run your workers locally. What they're currently working on in using

workerd

in miniflare to make remote and local dev work basically exactly the same.

I'm not sure if this is the right place, but I saw in lots of articles this week a request for libs that aren't yet supported on Workers to be posted, so I'd like to suggest https://github.com/slackapi/bolt-js which is Slack's SDK for building Slack Bots. I've built multiple Discord bots on Cloudflare and it's worked perfectly, but there's no equivalent library for Slack that works as far as I know. I'm assuming this lib doesn't work as it's not on the list of ones that work on Cloudflare. Secondly, and I don't know if this exists already, it'd be great if we could see what % of Node libraries or even just a page showing ticks and crosses are supported by Workers. I know there's some bullet points of which ones do, but be interesting to see how many of them are supported vs not and the gradual move towards (hopefully) 100%, or at least close, as I do think the main issue I come across is a lib not working on CF and it making life difficult. I think until that's better, it'd be a hard sell in my day job for example, either that or we'd likely use a WASM-supported library instead to get around the issue of libs not working.

An API that gets general information or metadata of individual durable objects is appreciated, for example: current size of a DO and how many keys there are; even just a reasonably close approximation. This could be very useful for SaaS analytics and metering, I find myself currently having to manually up and down the size variable every time storage size changes.

The ability to pass through a request to the origin server.

`performance.now`: we have a feature in our RPC framework to populate the

Server-Timing

header to give insight into each part of the request/response lifecycle, but because there is no way to time methods in a worker, the functionality doesn’t work; if you want to prevent timing attacks skew the resolution like other platforms, but there should be a means to tell how long CPU bound task take